Article
Automation of Lung Ultrasound Interpretation via Deep
Learning for the Classification of Normal versus Abnormal
Lung Parenchyma: A Multicenter Study
Robert Arntfield
1,
*, Derek Wu
2
, Jared Tschirhart
2
, Blake VanBerlo
3
, Alex Ford
4
, Jordan Ho
2
,
Joseph McCauley
5
, Benjamin Wu
6
, Jason Deglint
7
, Rushil Chaudhary
2
, Chintan Dave
1
, Bennett VanBerlo
8
,
John Basmaji
1
and Scott Millington
9
Citation: Arntfield, R.; Wu, D.;
Tschirhart, J.; VanBerlo, B.; Ford, A.;
Ho, J.; McCauley, J.; Wu, B.; Deglint,
J.; Chaudhary, R.; et al. Automation of
Lung Ultrasound Interpretation via
Deep Learning for the Classification
of Normal versus Abnormal Lung
Parenchyma: A Multicenter Study.
Diagnostics 2021, 11, 2049. https://
doi.org/10.3390/diagnostics11112049
Academic Editors: Keun Ho Ryu and
Nipon Theera-Umpon
Received: 14 October 2021
Accepted: 31 October 2021
Published: 4 November 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
1
Division of Critical Care Medicine, Western University, London, ON N6A 5C1, Canada;
Chintan.Dave@lhsc.on.ca (C.D.); john.basmaji@lhsc.on.ca (J.B.)
2
Schulich School of Medicine and Dentistry, Western University, London, ON N6A 5C1, Canada;
dwu2021@meds.uwo.ca (D.W.); jtschirhart2024@meds.uwo.ca (J.T.); jho2021@meds.uwo.ca (J.H.);
rchaud@uwo.ca (R.C.)
3
Faculty of Mathematics, University of Waterloo, Waterloo, ON N2L 3G1, Canada;
bvanberlo2021@meds.uwo.ca
4
Independent Researcher, London, ON N6A 1L8, Canada; aford3532@gmail.com
5
Faculty of Engineering, University of Waterloo, Waterloo, ON N2L 3G1, Canada; joewmccauley@gmail.com
6
Independent Researcher, London, ON N6C 4P9, Canada; benjamin.dq.wu@gmail.com
7
Faculty of Systems Design Engineering, University of Waterloo, Waterloo, ON N2L 3G1, Canada;
jason.deglint.engr@gmail.com
8
Faculty of Engineering, University of Western Ontario, London, ON N6A 5C1, Canada;
bennettjlvb@gmail.com
9
Department of Critical Care Medicine, University of Ottawa, Ottawa, ON K1N 6N5, Canada;
scottjmillington@gmail.com
* Correspondence: robert.arntfield@lhsc.on.ca; Tel.: +1-519-685-8786; Fax: +1-519-685-8089
Abstract:
Lung ultrasound (LUS) is an accurate thoracic imaging technique distinguished by its
handheld size, low-cost, and lack of radiation. User dependence and poor access to training have
limited the impact and dissemination of LUS outside of acute care hospital environments. Automated
interpretation of LUS using deep learning can overcome these barriers by increasing accuracy while
allowing point-of-care use by non-experts. In this multicenter study, we seek to automate the clinically
vital distinction between A line (normal parenchyma) and B line (abnormal parenchyma) on LUS by
training a customized neural network using 272,891 labelled LUS images. After external validation
on 23,393 frames, pragmatic clinical application at the clip level was performed on 1162 videos. The
trained classifier demonstrated an area under the receiver operating curve (AUC) of 0.96 (
±
0.02)
through 10-fold cross-validation on local frames and an AUC of 0.93 on the external validation
dataset. Clip-level inference yielded sensitivities and specificities of 90% and 92% (local) and 83% and
82% (external), respectively, for detecting the B line pattern. This study demonstrates accurate deep-
learning-enabled LUS interpretation between normal and abnormal lung parenchyma on ultrasound
frames while rendering diagnostically important sensitivity and specificity at the video clip level.
Keywords: deep learning; ultrasound; lung ultrasound; artificial intelligence; automation; imaging
1. Introduction
Lung ultrasound (LUS) is a versatile thoracic imaging method that offers the diag-
nostic accuracy of a CT scan for many common clinical findings, with all the advantages
of portable, handheld technology [
1
–
4
]. Since recent reports have highlighted that the
potential for LUS dissemination is near-limitless, for example, primary care, community
settings, developing countries, and outer space [
5
–
7
], accordingly, it has been praised as a
worthy upgrade to auscultation [
8
]. With experts in its use in persistent short supply [
9
–
12
],
Diagnostics 2021, 11, 2049. https://doi.org/10.3390/diagnostics11112049 https://www.mdpi.com/journal/diagnostics
